ITRI SEMINAR
               Thursday 13th November at 12.00


        A corpus-based study of anaphora in dialogues
                  in English and Portuguese

                        Marco Rochas
                    University of Sussex


The talk  describes  a  study of  anaphora  in  English  and
Portuguese face-to-face conversations.  The sources of  data
were the London-Lund Corpus, for English dialogues, and  the
Rio de  Janeiro  Corpus  of Clinical  Dialogues,  which  was
collected during the research  process. The approach  relies
on the manual annotation of a significant number of anaphora
cases - around three thousand for each language -  according
to four properties which will be described in the talk. Once
the  required   number  of   cases  had   been   analysed, a
probabilistic model was built  by linking categories  across
the four  properties to  form a  probability tree  based  on
aggregate co-occurrence. The  results were summed  up in  an
antecedent-likelihood  theory,  which   elaborates  on   the
probabilities and  observed  regularities of  the  immediate
context to describe anaphoric phenomena by means of patterns
for recognition  and  resolution. The  theory  was  manually
tested on a  previously annotated dialogue.  The process  of
statistical analysis and  the results will  be discussed  in
the talk. The same process  was repeated for the  Portuguese
data,   generating    a   probabilistic    model   and    an
antecedent-likelihood theory with distinctive features which
will be analysed  contrastively. The  talk concludes  with a
discussion on  possible  developments for  natural  language
processing.